Posted by
Glen MacDonald-2 on
Feb 18, 2009; 7:42pm
URL: http://imagej.273.s1.nabble.com/Re-IMAGEJ-Digest-14-Feb-2009-to-15-Feb-2009-2009-45-tp3693650p3693655.html
Gabriel,
thresholding was not the issue. The fish were thresholded, but some
were not selected by the particle analyzer.
Wayne looked at the stack and macro, and found that increasing the
circularity setting from .2 to .6 resulted in all being selected.
This also selected intersecting profiles of 2 or more, which is why I
was using a stringent value.
However, measurements of individual fish that had been missed did not
demonstrate circularity or area as the problems, and analysis on
single planes from the stack did not miss any fish.
today, I repeated the particle analysis with same parameters for
thresholding and particle analysis. it all works with no missed
objects.
In the end, we are checking our macro set for errant variables.
Regards,
Glen
On Feb 18, 2009, at 12:56 AM, Gabriel Landini wrote:
> On Wednesday 18 February 2009, Glen MacDonald wrote:
>> Analyze Particles correctly identifies about 25 larvae in each frame
>> of the timelapse.
>> But, 3-5 larvae are not measured in each frame.
>
> Unless you post pictures with examples and code, knowing what is
> going on will
> be just guesswork.
>
>> The display of the best fit ellipses and numbering the thresholded
>> larvae by result line identifies the missed larvae. Individual
>> measurement of the missed larvae indicates they meet all criteria for
>> particle analysis. In some instances, a larva may not have changed
>> position for several frames, yet will have been measured in some
>> frames, but not in others.
>
> There is a tendency to using only the built in threshold for object
> segmentation. While it works for many instances, it is by no means
> the optimal
> nor the most appropriate method for all possible cases. Perhaps you
> colleague
> should try other segmentation algorithms.
>
> G.